By Ben Fu and Justin Ouyang
As we wrap up an extraordinary year in tech, one thing is clear: the pace of innovation has never been faster. 2024 delivered groundbreaking advancements in AI products, adoption, and investment—reshaping industries and challenging our assumptions about what’s possible.
In our specific areas of focus, which includes AI, data, dev tools, and infrastructure, there’s been considerable progress, even though our efforts here only really began in earnest through the back half of the year.
Already, we’ve closed on two exciting new investments: an AI data platform, and an AI observability and evaluation platform—stay tuned for the formal announcements.
Looking ahead to 2025, here’s what we believe is on the horizon:
#1 OpenAI Takes the Crown
Forget the chatter about model commoditization and performance parity—OpenAI cements itself as the undisputed leader of this AI era. With the fastest-growing consumer app in history and surging enterprise adoption, OpenAI has set the record for pace for innovation. Their recent 12 Days of OpenAI reminded us all of their ability to ship rapidly and push boundaries across performance, modalities, and user experiences.
Despite the drama of 2024, OpenAI has stabilized and remains a magnet for top talent. In 2025, they’ll widen the gap, cementing their lead in both consumer and enterprise AI.
#2 The AI Reckoning: Fraud and Failure Emerge
The hype cycle always comes with casualties, and AI is no exception. We predict at least two high-profile AI frauds or failures will be exposed in 2025, mirroring the meteoric rise and fall of Stability AI.
The combination of quick term sheets, limited diligence, and sky-high valuations has created an environment ripe for overpromising and under delivering. As competition intensifies, some high-flying startups will falter under the weight of technical shortcomings, market pressures, or outright misrepresentation.
#3 Small Models, Big Impact, On Edge
2025 will see the rise of small language models (SLMs) and small specialized models (SSMs). With advancements in hardware, costs, and model optimization, these models will enable AI to thrive on-device, unlocking use cases that were previously limited by cloud dependence.
Expect big tech players like Google and Apple to lead the charge, with cracks beginning to show in Nvidia’s dominance as AI shifts from large-scale training to edge applications.
#4 On-Prem is Back
The cloud revolution was transformative, but the pendulum may be swinging back. For enterprises managing sensitive data or seeking greater control, on-premise infrastructure is becoming attractive again—particularly for AI projects with unique performance or security requirements.
While hybrid approaches will dominate, 2025 could mark the end of the cloud migration as the default strategy. Our current cloud transition state may begin to look a lot more like the end state.
#5 The First AI-Native Security Breach
The rise of agentic AI introduces new vulnerabilities, and 2025 could see the first AI-native cybersecurity breach—a malicious attack developed and executed with AI technologies. The barrier to develop malware or malicious attacks has been lowered. We will see both more volume and more sophistication.
From deepfake phishing to AI-powered malware, the sophistication of attacks will escalate. This will spark a new era of AI-native cybersecurity solutions and even geopolitical AI weapons races.
#6 Scaling Laws Hit a Wall
The exponential improvement of AI models may slow as we hit the limits of more data and more compute. We’ll see diminishing returns until new breakthroughs in training techniques, architectures, or the underlying resource equation emerge.
In 2025, the focus will shift to tuning and optimization, as well as exploring orthogonal approaches that could redefine the AI landscape. We suspect many of these developments are already underway within some of the leading foundational AI companies.
#7 Crypto x AI
As the AI hype cools, we’ll see opportunistic founders migrate back to crypto, chasing the next big funding wave. But this time, expect the convergence of AI and crypto to yield intriguing concepts like decentralized compute, agent wallets, or tokenized energy markets.
#8 Leadership Shakeups at Hyperscalers
The race to dominate AI has raised the stakes for hyperscalers. Incumbents are competing better than ever, and the pressure is mounting. We predict at least one hyperscaler CEO will face the axe in 2025 for failing to adequately deliver on AI promises.
#9 The Great Venture Capital Culling Continues
The VC landscape is undergoing a transformation. Fewer LP dollars are flowing into venture funds, and firms are finding it harder to raise. The result? Fewer products, fewer seats, and only the best brands and strategies surviving.
The days of “one fund fits all” are over. Expect more focus and discipline in fund strategies—and a challenging market for firms stuck in the middle.
Bonus: VC Nuclear Hype Fades
Nuclear power, while a vital piece of the clean energy puzzle, isn’t a natural fit for venture capital. With its long cycles, high upfront costs, and regulatory barriers, nuclear will remain better suited for infrastructure investors.
Yes, it’s a stable, durable, and sizable source of “clean” energy. But we don’t see a future where political powers allow a proliferation of reactors – mini or not. But there are opportunities for startups to streamline the manufacturing and operations of these facilities.
These are just a few of the trends and dynamics we’re watching as we head into 2025. We’d love to hear your thoughts—what are you predicting for the year ahead? Reach out and let us know!
